# AI for local government

> In local government, an AI agent is most useful answering resident questions from information the agency has already published, structuring permit and service request intake, finding records for staff, and preparing meeting materials from existing documents. It answers only from approved sources, cites them, and routes anything personal, contested or unpublished to a person. Accessibility, transparency and public records obligations shape the design, so those questions go to the clerk and counsel first.

Source: https://risendigital.io/industries/ai-for-local-government  
Publisher: Risen Digital  
Updated: 2026-09-21

## At a glance

- **Best first use:** Resident questions that your website already answers, because the source is public and the answer can be cited.
- **What it reads:** Published pages, ordinances and fee schedules, forms, meeting agendas and minutes, and internal records it has been explicitly permitted to see.
- **What it writes:** Draft answers, structured intake records, records search results with sources, and meeting briefing drafts, held for staff review.
- **Where people stay:** Decisions on applications, enforcement, anything involving an individual's personal information, and any answer not found in an approved source.

Residents ask the same questions every day, permits and service requests arrive in every channel, and staff spend hours finding records and preparing for meetings. This page shows where an AI agent can help a city, county or agency with that work, what it must only ever answer from, and which questions to put to your clerk and counsel before you start.

![The front of a city hall building on a spring day](https://risendigital.io/images/experience/city-hall-1600.jpg)

## What can AI do for a city, county or public agency?

It answers routine resident questions from published information, structures the intake of permits and requests, helps staff find records, and prepares meeting materials from documents the agency already holds. Decisions stay with staff and elected officials.

Public agencies hold a great deal of information and are asked about it constantly. When is bulk pickup, what does a fence permit require, how do I contest a parking ticket, is the pool open on Sunday. The answers are on the website, in an ordinance, on a fee schedule or in a form's instructions. What is missing is a way for a resident to ask in their own words and get the right paragraph back, at any hour, in their language, without a phone queue.

The same gap exists inside the building. A council member asks what was decided about a rezoning years ago; a clerk searches minutes. A permit technician needs a fee; they open a PDF. An AI agent grounded in the agency's own documents reads them once and lets people ask. It does not decide anything and it does not invent anything; it finds, structures and drafts.

| Task | What the agent does | Who decides |
| --- | --- | --- |
| Resident questions | Answers from published pages and documents, with the source linked | Staff, for anything not covered or personal |
| Permit and request intake | Collects the required details, checks for missing items, creates the record in the right system | The reviewing department |
| Records lookup | Searches minutes, ordinances, resolutions and files staff may access, and cites results | Staff judge relevance and handle any release |
| Meeting preparation | Drafts briefing summaries and background from agenda items and prior decisions | The clerk and the responsible department |

## Where is your agency ready to use AI, and where is it not?

Work through the checklist by area. Where most boxes are already ticked, an agent can start in draft mode; where they are not, the items themselves are the preparation work.

_Interactive on the web page: Local government AI readiness checklist. The checklist covers five areas. Resident-facing services: whether the website is current, whether common questions have written answers, whether service channels are known and whether accessibility standards are applied. Internal operations: whether intake steps are documented, which systems hold requests, and whether staff have time to review. Data and records: whether records are searchable, whether retention schedules are known and whether personal information is separated from public information. Procurement and governance: whether the purchasing route, the responsible official and the review process are identified. Staff and training: whether staff know what the agent can and cannot do and how to escalate. Agencies that tick most items in the resident-facing and records areas are usually ready to begin with published-information questions in draft mode; the governance and training items are what let a pilot become a service._

## How does an AI agent answer resident questions without guessing?

It is restricted to an approved set of sources, retrieves the relevant passage before answering, shows where the answer came from, and says when it has not found one. Guessing is prevented by design, not by hoping the model behaves.

1. **The agency chooses the sources** Published web pages, ordinances, fee schedules, forms and their instructions, agendas and minutes. Nothing else is in scope, and a department owns the list.
2. **The resident asks in their own words** On the website, by text message or by phone, in the languages the agency supports.
3. **The agent retrieves before it answers** It searches the approved sources for the relevant passages and composes an answer from them, with a link to each.
4. **It answers, or it says it cannot** If no source covers the question, the agent says so and offers the right department, a form or a person. It does not fill the gap from general knowledge.
5. **Personal matters go to a person** A question about a specific account, ticket, application or address is handed to staff with the conversation attached, because answering it means touching personal information.

Permit and request intake works the same way in the other direction. The agent asks the resident for what your documented process requires, checks that nothing is missing, and creates the record in the system the department already uses. It does not assess the application. The technician sees a complete record, and the resident learns what is missing before they submit rather than a week later.

## Can AI help staff find records and prepare for meetings?

Yes. An agent grounded in minutes, resolutions, ordinances and files that staff are permitted to see can answer questions about past decisions with citations and draft briefing material for agenda items. Staff judge what is relevant and what is released.

Records lookup is where staff time disappears quietly. Finding when a matter was last before the council, what was decided and what conditions were attached can take an afternoon across minutes, packets and email. An agent that has indexed those documents answers in a minute, with the meeting date and the page. The person still reads the source; they just do not have to find it.

Meeting preparation builds on the same index. For each agenda item the agent drafts a short background: the history of the matter, prior decisions, related ordinances and open questions from the last discussion. The clerk and the department edit it, and members arrive having read a short summary with the full packet still there when they need it.

> **Internal does not mean private** Drafts, prompts and search logs produced inside the agency may be public records. Ask the clerk how they should be retained and whether they are releasable before the first internal use, not after the first request.

## What should you check about accessibility, transparency and public records?

Treat these as questions for your clerk, your counsel and your accessibility lead. The agent can be built to meet almost any answer, but the answers belong to the agency and its obligations, not to the vendor.

- **Accessibility** Which digital accessibility standard does the agency apply, and has the interface been tested against it with real assistive technology? Is a non-digital route always available?
- **Transparency** Will residents be told they are talking to an automated system? Will the sources of each answer be visible? Who is named as responsible for the service?
- **Public records** Are conversation logs, intake records and internal drafts records under your retention rules? How long are they kept, where, and how would they be produced if requested?
- **Personal information** What personal data could the agent see, where is it processed, and does that fit the agency's privacy policies and any applicable rules?
- **Procurement** Which purchasing route applies, what vendor terms are required, and does the model provider's contract meet them?

> **This page describes questions, not requirements** Obligations differ by jurisdiction and by type of agency. Nothing here is legal guidance. Put each question to the clerk, counsel or the official responsible, and record their answers as part of the project's design.

## How does an agency start responsibly?

Begin with published-information questions in draft mode, with staff reviewing every answer, the sources visible and the governance questions answered in writing, then extend to intake and internal records on the strength of the record.

1. **Answer the governance questions first** A short written document from the clerk, counsel and the accessibility lead. It prevents the pilot being stopped later.
2. **Choose the top resident questions** The ones the front desk answers most. Check the published answers are correct and current, because the agent will repeat them.
3. **Run in draft mode internally** Staff ask the agent the questions residents ask and grade the answers. Fix the sources where answers are wrong.
4. **Open one channel to residents** With the automated nature disclosed, sources shown, and a person one step away. Review the logs weekly.
5. **Extend by evidence** Add intake for one permit type, then records lookup for one department. Each step has its own review and its own record.

## Frequently asked questions

**Q: Can an AI agent make decisions on permits or applications?**  
A: No, and it should be built so that it cannot. The agent collects and structures the information a documented process requires, checks completeness and creates the record. Assessment, approval, denial and any enforcement step are decisions for staff and, where required, for the governing body. That boundary is written into the agent's permissions.

**Q: What stops the agent giving residents wrong information?**  
A: It answers only from sources the agency has approved, retrieves the relevant passage before composing an answer, and links to it so the resident and staff can check. When no source covers the question it says so rather than guessing. Errors then trace back to a published source, which has an owner who can correct it.

**Q: Are chat logs with residents public records?**  
A: That is a question for your clerk and counsel, and the answer may differ by jurisdiction and by what the logs contain. Ask before the first conversation happens, decide how logs are retained and how they would be produced, and design the agent so that personal information is kept out of them wherever possible.

**Q: How do we keep the service accessible to everyone?**  
A: Test the interface with the assistive technology residents actually use, apply the accessibility standard your agency follows, write answers in plain language, support the languages your community speaks from checked sources, and keep the phone, the counter and the form as full alternatives. An agent adds a channel; it should never remove one.

## Related pages

- [Conversational AI for business](https://risendigital.io/guides/conversational-ai-for-business)
- [AI data privacy and security questions](https://risendigital.io/guides/ai-data-privacy-and-security-questions)
- [AI readiness checklist](https://risendigital.io/guides/ai-readiness-checklist)
- [What is an AI agent?](https://risendigital.io/guides/what-is-an-ai-agent)